Prior mental fatigue reliably makes endurance exercise feel harder, but its effect on performance is probably small and remains contested
Interpretation
Early work found large effects: 16 participants cycled to exhaustion about 15% sooner after 90 min of demanding cognitive work, with higher perceived effort and unchanged physiology (Marcora 2009), and reviews found consistent declines (Van Cutsem 2017, 11 studies; Brown 2020, 73 studies, 2,581 participants, aerobic g = -0.26). But bias-sensitive analyses shrink the effect to near zero (Holgado 2020, corrected 2021: dz = -0.53 uncorrected, -0.14 bias-corrected, non-significant; Holgado 2023: median power 39%), pre-registered experiments in 30 and 22 recreational adults found no performance cost (Holgado 2021, 2023), and the largest trial (117 adults, 57 women) found only a non-significant trend in 20-min time-trial distance (d = 0.20) alongside higher RPE and slower reaction time (Habay 2026). Higher perceived effort is the consistent part; the performance effect is where studies disagree. Lab computer tasks, mostly recreational adults, no triathletes.
Practical implication
Expect sessions to feel harder after mentally draining days and don't read one poor session as lost fitness; use objective targets, move key sessions to lighter days where possible, and take extra care riding outdoors when attention is low. Extrapolated from lab studies in recreational adults, not triathletes or real working days; persistent exhaustion needs medical review.
Sources
9 sources, independently verified.
After 90 min of a demanding cognitive task vs 90 min of neutral documentaries, cycling time to exhaustion at 80% of peak power fell from 754 to 640 s (about 15%). Cardiorespiratory and metabolic responses and motivation were largely unaffected; perceived effort was higher, so participants reached maximal effort sooner. Small single study; not replicated in a pre-registered replication (Holgado 2021).
Mental fatigue was associated with reduced endurance performance (shorter time to exhaustion, lower self-selected power or speed) and higher perceived exertion; heart rate, lactate, oxygen uptake, cardiac output and VO2max unaffected; maximal strength, power and anaerobic work unaffected. No bias correction.
Prior cognitive exertion reduced physical performance, g = -0.38 (95% CI -0.46 to -0.31); aerobic g = -0.26, isometric and motor -0.57, dynamic resistance -0.51, maximal anaerobic trivial (0.10, ns). Between-subject designs g = -0.65 vs within-subject -0.28. Mostly small lab studies in non-athletes.
After the authors' 2021 correction (doi:10.5334/joc.178, miscoded effect-size signs fixed, two RPE measures added): pooled effect of mental fatigue on endurance performance dz = -0.53 (95% CI -0.76 to -0.28; originally -0.50) and on RPE dz = -0.32. Egger's test and a three-parameter selection model indicated publication or reporting bias; bias-corrected performance effect -0.14 (95% CI -0.46 to 0.16), non-significant. Conclusions unchanged: evidence does not conclusively show that mental fatigue harms exercise performance.
Traditional meta-analysis dz = -0.54, but bias-adjusted estimate essentially zero (dz = 0.01, inconclusive, BF10 = 0.90). Median observed statistical power 39% for the unadjusted effect and 20% for a more modest one: most studies were underpowered.
Replication of the Marcora design: the 90-min task increased subjective mental fatigue, but there was moderate-to-strong Bayesian evidence for no reduction in time to exhaustion at 80% VO2max (BF01 = 9.762).
An individualised mental-effort task raised subjective mental fatigue, but time to failure at 80% of peak power was similar (control 410 s vs mental fatigue 422 s, BF10 = 0.15) and knee-extensor maximal force was not impaired.
45-min individualised Stroop task vs self-chosen documentary. GoNoGo reaction time worsened (P < 0.001); 20-min cycling time-trial distance showed only a non-significant trend (P = 0.059, d = 0.20) with slightly lower cadence; feelings of physical and mental fatigue and RPE stayed higher during the time trial.
After a 30-min Stroop task vs a 10-min easy task, recreational cyclists' 20-min time-trial power fell (216 vs 226 W) while professionals' did not (323 vs 326 W); professionals had better Stroop performance. Small cross-sectional comparison; cannot show training causes resistance.